Because electronic parts have been increasing their heat generation as their integration, volume, output, etc. are increasing, demand is mounting on materials having a high thermal conductivity and a small thermal expansion coefficient. Because semiconductor devices such as CPUs, light-emitting diodes, etc. generate large amounts of heat, heat sinks are usually mounted to them. Heat transmitted from the semiconductor devices to the heat sinks is dissipated by fans or cooling media, etc. The heat sinks are usually made of aluminum, copper or their alloys having excellent thermal conductivity.
Because CPU, for instance, is much smaller than the heat sink, a high-thermal-conductivity body called “heat spreader” is usually interposed therebetween. Materials for the heat spreader preferably have a high thermal conductivity and as small a thermal expansion coefficient as that of the CPU made of silicon. This is true of light-emitting diodes made of compound semiconductors (GaAs, GaN, etc.). For such purposes, a lot of substrates made of composite materials of ceramics having small thermal expansion coefficients such as silicon carbide, alumina, silicon nitride or aluminum nitride and aluminum or copper have been proposed. However, substrates made of these composite materials are disadvantageously poor in workability because of containing ceramics. Though composite material substrates made of metals having small thermal expansion coefficients such as tungsten or molybdenum and copper have also been proposed, these composite material substrates are disadvantageously poor in workability.
Under the above circumstances, a lot of attempts have recently been proposed to use composite materials of carbon particles or fibers and metals for heat-dissipating substrates. For instance, JP 10-168502 A discloses a high-thermal-conductivity composite material obtained by mixing 1 to 200 parts by weight of one or more crystalline carbon materials selected from the group consisting of graphite, carbon fibers, carbon black, fullerene and carbon nanotubes, and 100 parts by weight of metal powder selected from the group consisting of Fe, Cu, Al, Ag, Be, Mg, W, Ni, Mo, Si, Zn and these alloys and hot-pressing the resultant mixture. However, because this composite material has a structure containing crystalline carbon materials dispersed in a metal matrix, it has as large a thermal expansion coefficient as that of the metal matrix, though it has a high thermal conductivity.
JP 2000-203973 A discloses a carbon-based metal composite material comprising a carbonaceous matrix impregnated with at least one metal selected from the group consisting of aluminum, magnesium, tin, zinc, copper, silver, iron, nickel and their alloys, 90% by volume or more of voids in the carbonaceous matrix being impregnated with the metal, and the amount of the metal being 35% by volume or less of the entire carbon-based metal composite material.
JP 2001-58255 A discloses a carbon-based metal composite material produced by impregnating carbon moldings comprising carbon particles or fibers containing graphite crystals with aluminum, copper, silver or these alloys at high pressure by a melt-forging method, which has a thermal conductivity of 150 W/mK or more and a thermal expansion coefficient of 4×10−6/K to 12×10−6/K in a thickness direction at room temperature. This carbon-based metal composite material has a structure comprising a graphite matrix having high rigidity, a high thermal conductivity and a small thermal expansion coefficient as a skeleton, with its voids impregnated with a metal. Accordingly, it has both small thermal expansion coefficient inherent in graphite and high thermal conductivity inherent in a metal.
Despite the above advantages, these carbon-based metal composite materials are disadvantageous in having much larger thermal expansion coefficients than those of silicon and compound semiconductors. Large differences from silicon and compound semiconductors in a thermal expansion coefficient undesirably exert large thermal stress to CPUs or light-emitting diodes during soldering or brazing heat-dissipating substrates to the CPUs or the light-emitting diodes, or during the operation of the CPUs or the light-emitting diodes. Accordingly, a stress-relieving member is usually interposed between the CPU or the light-emitting diode and the heat-dissipating substrate. However, because the stress-relieving member does not necessarily have a sufficiently large thermal conductivity, the use of the carbon-based metal composite material having a high thermal conductivity for the heat-dissipating substrate does not exhibit its effect sufficiently.
Further, when the heat-dissipating substrate is bonded to the CPU or the light-emitting diode, soldering is usually conducted at about 200 to 300° C. in the case of an aluminum-impregnated graphite substrate, and brazing is usually conducted at about 700 to 800° C. in the case of a copper-impregnated graphite substrate. It has been found, however, that when exposed to such high temperature, the composite substrate impregnated with aluminum or copper exhibits extremely different sizes due to a residual stress before and after the heating. Such thermal hysteresis causes warp in the heat-dissipating substrate bonded to the CPU or the light-emitting diode, so that the heat-dissipating substrate may finally be broken, or that the CPU or the laser diode, etc. may also be damaged by thermal stress.